new advances in passive seismic tomography from ...new advances in passive seismic tomography from...

New advances in passive seismic tomography from instrumentation to data processing andinterpretation – some examples

A. Tselentis* Greek Geodynamic Institute, S. Kapotas Landtech Group, N. Germenis LandtechGeophysics, N. Russil, Landtech Geophysics,

[email protected]

Passive, Tomography, Earthquakes

Summary

Passive Seismic Tomography (PST) in recent years hasbecome a well established technique in the upstreamhydrocarbon industry specifically for regional hydrocarbonexploration. Significant advantages have been achieved notonly in instrumentation but in data de-noising, processingand interpretation. Some of these advantages will bepresented with case histories.

Introduction

PST has been applied successfully in regional hydrocarbonexploration, demonstrating its potential to map large areasfor a relatively low cost compared to conventional 1Dseismic surveys (Kapotas et al., 2003). Valoroso et al.(2008) use 4D passive seismic tomography to detect space-time dependency in response to fluid pressure. Tselentis etal. (2006) show that PST can even be applied at a localscale. Although the number of such applications is limitedat the moment, with improvements in instrumentation, dataacquisition and processing technology, the use of PST as atool for hydrocarbon exploration and characterization islikely to flourish.

The rationale for applying PST as a complementaryimaging tool has several important advantages. First,tomography is a cost-effective means of imaging a largearea with difficult terrain in which conventional seismicexploration is expensive and can be of poor quality becauseof seismic penetration problems. Second, PST can providean accurate 3D velocity model that can be used to improve(i.e., PSDM) existing or lower-quality reflection seismicdata. Third, the technique is environmentally friendly, animportant consideration in all operational activities.Finally, PST can provide parameters related directly to

reservoir properties, such as VP=VS and QP. Theseparameters are very difficult to derive from conventionalseismic techniques because they require large-amplitudeshear waves.

Processing of PST data at a local scale for hydrocarbon ex-ploration is more complicated than applying off-the-shelf3D inversion algorithms. To get the best resolution of thegeologic formations at the lowest cost, we tap an arsenal of

techniques, including initial-velocity model selection,simultaneous earthquake hypocenter and 3D velocitymodels, QP inversion, synthetic and real-data checkerboardtests.

Data de-noising and phase identification

During a PST investigation, we use small magnitudeearthquakes which often are hard to detect since they canbe corrupted by noise. Therefor, identification of the weakP- and S-wave arrivals is important for obtaining accuratePST results.

For high resolution PST applications, we need as many aspossible small magnitude events which can becharacterized as point sources. These small events,especially if acquired in urban areas are often stronglyaffected by noise, so we also require procedures that allowa reliable first arrival picking, without losing importantinformation. During the last years various methodologieshave been developed.

Thus an important step in the processing is to attenuate theenergy of the seismic noise while not only preserving theenergy from the event but also avoiding altering the arrivaltimes of the seismic phase. A number of algorithms havebeen proposed for automatic event detection and seismicphase identification and can be divided in the followingcategories:

Automatic Microseismic Event Detection

i. Energy Based Algorithm (recommended for highseismicity records)

• Improved STA/LTA algorithm.• Dynamic threshold based on the statistical properties

of the STA/LTA “ratio”.• Simple and fast, demands low computational

resources.

ii. Algorithm Based on Statistical Methods(recommended for noisy records)

Two – stage procedure, based on a non-strict hypothesistesting scenario:

11th Biennial International Conference & Exposition

Advances in passive seismic tomography

• First stage: Estimation of the empirical pdf of theseismic noise (using statistical methods such assampling, modeling, clustering).

• Second stage: Use of a thresholding scheme inorder to detect the microseismic events, in a non-strict hypothesis testing framework.

iii. Algorithm Based on Signal’s PolarizationAttributes in Time-Frequency domain(recommended for extremely noisy records)

• Fourier analysis on different frequency sub-zones.• Evaluation of the polarization differences among the

three components.• Regression analysis technique in order to correct

errors due to sensor’s interference.• Development of a characteristic function based on

the above differences, for the microseismic eventdetection.

Automatic P&S Phase detection

i. P-phase picking:• The kurtosis criterion is applied on the segment of

the record that includes a seismic event.• The maximum slope of the kurtosis curve is

assigned to the P-onset time.ii. S-phase picking:• Eigenvalue analysis on 3C data.• Development of a characteristic function, based on

the maximum eigenvalues of the above analysis.• Kurtosis criterion on the characteristic function

Recently LandTech developed an integrated Algorithm(AYTOPSI©) for seismic events detection, signalenhancement and automatic P- and S-phase picking. Thismethod is comprised by a Chi-squared based test statisticaltest for the event detection, filtering in the S- transformdomain, for de-noising and an automatic picker based onthe Kurtosis criterion. The performance of the method hasbeen applied successfully to various datasets and seems tosolve the problem of automatic processing of seismologicaldata of large PST networks. Automatic picks werecompared against manual reference picks, resulting in meanresidual time of 0.051 s. Moreover, this new technique wassubjected to a noise robustness test and sustained a goodperformance. The mean residual time remained lower than0.1 s, for noise levels between −1 up to 8 dB. Figure1depicts the various stages of this method

Passive instrumentationThe magnitude (Richter Scale) of the recorded eventsduring a PST survey, range from –2R to 3R. For therecording of such small events, the sensor must be veryhaving sensitivity at least 1500V/m/sec.

Figure 1: AYTOPSI algorithm.

Figure 2: Zoomed area shows the event selected to applythe methodology. Vectors indicate the detected events andthe red dots indicate the presence of the seismic events.

Figure 3: The event selected from (a)the real data and (b)the filtering in the time-frequency domain. The red dashedline is the automatic pick as calculated using the kurtosiscriterion and the black doted line is the actual position ofthe first arrival.



The resolution of the digitizer is one of the most importantparameters of the seismic instrumentation. 24bits digitizersare mostly used in seismotectonic studies and rot suitablefor PST investigations. These surveys require 4rd

generation 32bit digitizers, with dynamic range at least138dB at 250sps and 142dB at 1000sps

(a)

(b)

Figure 4: (a) Noise spectrum of the Sr32 LandTechdigitizer, (b) corresponding noise histogram.Another issue is the autonomy of the PST seismic stations.Given that are powered from a simple 12V lead acidbattery, the power cycle must be as many days as theseismic crew needs to visit the seismic station. Seismicnetworks consisted of 150 seismic stations, spread in anarea of 3000 – 5000 Sqm2 placed in accessible terrain,usually takes two weeks to twenty days from the seismiccrew to visit them. So the power autonomy of each seismicstation (recorder + sensor) must be enough for at least 20-25 days. There should be also enough storage capacity forstore the data of the above period.

Figure 5: Photo of a new generation 32bit PST seismographdeveloped by Landtech-Geophysics and Geobit. It ispowered by a motorcycle battery for more than a month.

The seismic events obtained during a PST explorationusually have frequency spectrum into the band 2Hz to20Hz. So a wide band seismometer with the range of below1Hz up to 30 Hz is necessary. Ideally PST sensors withbandwidth from 0.1Hz to 98Hz are preferred in order torecord these seismic events with the maximum quality. Thelow frequency response gives also the ability to calculatemoment tensors as well.PST sensors must be of borehole type, so they can beinstalled in a few meter depth, depending on noise. Thenoise level at this depth is much less than the surface. Theborehole can be easily (opened with a drilling machine withlow cost. Small diameter boreholes can be opened by handin rough areas where no vehicles can access.

(a)

(b)

Figure 6: (a) recording at the surface, (b) recording at 10mdepth.

ProcessingProcessing of PST data at a local scale for hydrocarbonexploration is more complicated than applying off-the-shelf3D inversion algorithms. To get the best resolution of thegeologic formations at the lowest cost, we tap an arsenal oftechniques, including initial-velocity model selection,simultaneous earthquake hypocenter and 3D velocitymodels, QP inversion, and synthetic and real-datacheckerboard tests.



Figure 7: The stages of PST data processing.

A hydrocarbon reservoir tends to be acoustically softer thanregions that are full of an incompressible fluid such aswater. Thus, a seismic wave should suffer more attenuationin a hydrocarbon reservoir than in surrounding materials.One of the geophysical parameters that correlates best withthe physical state of the rocks and the percentage of fluidcontent is the intrinsic quality factor QP of thecompressional body waves. As an exploratory tool,attenuation effects have only recently attracted attention(e.g., Tselentis et al., 2010). The QP can prove useful in twoways: as a means of correcting seismic data to enhanceresolution of conventional imaging techniques and as adirect hydrocarbon or geothermal indicator. Thereconstruction of QP imaging is considered to be a powerfultool for establishing the distribution of fractured systemscharacterized by fluid circulation.Until recently, most attempts to extract attenuation on alocal scale have been restricted to active seismic datarecorded at the surface. This approach encounterssignificant difficulties because the amplitude spectrum ofthe seismic record contains the imprint of the amplitudespectrum of the earth’s reflectivity as well as the amplitudespectrum of the seismic wavelet.A method based on the inversion of the rise times isexpected to provide the most reliable estimates of intrinsicattenuation. In fact, because only a very limited portion ofthe seismogram is used, the effects of multiple wavesgenerated in thin layers around the recording site areusually minimized (de Lorenzo et al., 2006).A mathematical model for realistic pulse broadening in anin homogeneous medium has been suggested by Gladwinand Stacey (1974) and Stacey et al. (1975). These studies

show experimentally that the rise times of acoustic signalspropagating linearly in elastic media with frequency-independent quality quotient Q is described by

where s is the pulse rise time, τ0 is the original pulse risetime at the source, C is a constant, ds is an arc segmentalong a ray-path, and ΔT is the incremental travel time.

Figure 8: Measurement of rise time and pulse widthbroadening.For a medium with constant VP, where Q = Q0, the aboveequation can be written in a linear form:

The ratio T=Q is usually referred as t* in the literature.The constant C was determined experimentally forultrasonic acoustic pulses to be 0.5 (Gladwin and Stacey,1974)

The following figure presents an example of a 3D mappingof Qp below an exploration block.

(a) (b)

(c)

Figure 9: (a) Seismic station layout, (b) 3D distribution ofrecorded hypocenters, (c) 3D Qp structure of theinvestigated area. Lower Qp regions have been removedrevealing the basement.



Evaporite identification

Evaporite identification from geophysical data is animportant task for oil prospecting in new regions where fewor no wells have been drilled. In this paper, we propose toapproach the problem of evaporite identification throughthe use of a two step procedure. Initially estimating seismicparameters such as P and S wave velocities and Poissonratio for the area of interest through the use of passiveseismic tomography and then using Kohonen neuralnetworks techniques to perform data clustering, patternrecognition and classification.

To further analyze the clustering of the data and reveal themajor lithological units in the region, we use Kohonenneural networks or self organizing maps (SOMs). These areunsupervised artificial neural networks developed byKohonen (1982), who intended to provide ordered featuremaps of input data after clustering. That is, SOMs arecapable of mapping high-dimensional similar input datainto clusters close to each other on a n- dimensional grid ofneurons (units).

Figure 10: Vp,Vs and Poisson’s ratio are used to “train” theneural network.

Figure 11: Example of 3D evaporate delineation below anentire exploration block. Not the match with the surfaceoutcrops.

A 3D delineation of the evaporitic cup of a gas reservoir inAlbania which was derived from a PST survey is depictedin the following picture.

Figure 12: Delineation of the evaporitic cap of a reservoirusing the 3D distribution of the Vp/Vs parameter.

Gas identification

Low VP/VS values characterize gas-bearing rocks (i.e.,those with a high fluid compressibility), whereas highervalues of VP/VS indicate liquid-bearing formations (i.e.,those with a low fluid compressibility). Furthermore, pore-fluid pressure may also play a role by inducing a fluid-phase transition and by keeping pores and cracks open. Asa consequence, velocities are further affected. Laboratorymeasurements show that crack opening induced byincreasing pore pressure leads to a strong reduction inVP/VS in gas-bearing rocks.

Figure 13. Relation between VP and VS, at the producingdepths 2204 of the gas (blue circles) and oil (red stars)fields.



The figure above, presents the relationship between VP andVS values, which were obtained from the passive surveybut only for the corresponding production depths of the gasand oil reservoir. Despite the overlapping region, there is atendency for the oil and gas values to separate from eachother on the cross-plot.

Figure 14: Successful delineation of water saturatedcarbonates, evaporates and gas reservoir from Vp/Vs valuesobtained during a PST survey.

Drilling sites

Although a PST survey provides mainly, structural andlithological information below the entire block of interestand cannot be considered as a direct hydrocarbon indicator,in certain cases we can provide maps that depict sub-regions within the investigated block that possess a highprobability for a successful well. This is due to the fact thatone of the products of PST is the Poisson's ratio, whichstrongly depends on the fluid content within the pore space.Furthermore, the mapping of the Q parameter all over anexploration block could reveal regions of high porosity,fracture zones and help differentiate gas and oil in the porespace.

Figure 15: Setting certain limits on the derived parameterswe can locate sites (+) possessing high probabilities for asuccessful well.

Conclusions

PST can provide a wealth of information for areas whereconventional 2D seismic surveys do not work. PST is themost appropriate method for regions with severe seismicpenetration problems and difficult topography as well asregions with environmental restrictions. Recent advantagesin instrumentation and processing methodologies havesignificantly increase the resolution of PST methodologywhich can now be applied at a reservoir scale. Some ofthese advances are presented in the present paper.

References

de Lorenzo, S., M. Filippucci, E. Giampiccolo, and D.Patane, 2006, Intrinsic Qp at Mt. Etna from the inversion ofrise times of 2002 micro- earthquake sequence: Annals ofGeophysics, 49, 1215–1234.Gladwin, M. T., and F. D. Stacey, 1974, Anelasticdegradation of acoustic pulses in rock: Physics of the Earthand Planetary Interiors, 8, no. 4, 332–336.Kapotas, S., G.-A. Tselentis, and N. Martakis, 2003, Casestudy in NW Greece of passive seismic tomography: A newtool for hydrocarbon exploration: First Break, 21, 37–42.Kohonen, T., 1995, Self-organizing maps: Springer, BerlinStacey, F. D., M. T. Gladwin, B. McKavanagh, A. T.Linde, and L. M.Hastie, 1975, Anelastic damping ofacoustic and seismic pulses :Surveys inGeophysics,2,2,133-151Tselentis, G., E. Sokos, N. Martakis, and A. Serpetsidaki,2006, Seismicity 2844 and seismotectonics in Epirus,western Greece: Results from a micro- 2845 earthquakesurvey: Bulletin of the Seismological Society of America,96, no. 5, 1706–1717.Tselentis, G., P. Paraskevopoulos, and N. Martakis, 2010,Intrinsic Qp 2841 seismic attenuation from the rise time ofmicroearthquakes: A local 2842 scale application at Rio-Antirrio, western Greece: GeophysicalProspecting,58,5,845-859.Valoroso, L., L. Improta, P. De Gori, R. Di Stefano, L.Chiaraluc, and C. Chiarabba, 2008, From 3D to 4D passiveseismic tomography — The sub-surface structure imagingof the Val d Agri region, S. Italy: 70th Annual EAGEConference & Exhibition, Extended Abstracts, P028

Acknowledgements

LandTech Enterprises and Stream Oil are acknowledgedfor providing permission to publish the results of thisinvestigation. Finally, we would like to thank the technicaland scientific personnel of LandTech Enterprises, withoutwhose contribution and professionalism this survey wouldnot have been a success.


new advances in passive seismic tomography from ...new advances in passive seismic tomography from...

Documents